Shulou(Shulou.com)06/02 Report--

In an area where technology decisions take years and continue to be used for decades, we have found that significantly reducing the time to achieve the first value can reduce the risk for technology developers and adopters, and more information is available in the vibrating chain.

We fundamentally reconsidered operating AI and how to reduce production value time from months to days to apply it to production operations.

Shorten the time to achieve the first value, so that plant and manufacturing executives who believe in the potential of operating AI can quickly prove their intuition, while the risk is much lower.

Patterns in operational data

An accepted premise is that operational data should contain evidence of pattern-system behavior, including early warning and root causes. This premise is the basis of remote monitoring, industrial Internet of things and predictable operational excellence.

Currently, implementing manufacturing digitization projects takes three to five years and tens of millions of dollars, and requires a large team of practitioners with three or more skill categories (data engineering, data science, software management). This is a high-risk job that may be feasible in an economy where it is easy to make money. However, this is unsustainable and often leads to difficulties in development after the pilot phase. Workbench and AirGap

Falkonry helps customers overcome these difficulties through its products (Workbench and AirGap versions) and extends by focusing on citizen data science. The existing analysis goal is to visualize shapes, waveforms and features, but only Falkonry automatically discovers important operational data features and explains their importance and importance. This is the core innovation in which we have obtained four patents. This is also the basis for the naming of CBInsights AI 100s in 2019 and 2020.

That's why the Air Force chose Falkonry as its strategic plan to extend its deployment of operating AI technology to other Department of Defense (DoD) customers. With Falkonry, manufacturers can develop and deploy production models within a year. A $10B + steel company has used Falkonry's Operational AI to achieve this result in dozens of use cases, 10 times faster than the alternative. Although this level of success is the best of its kind, it is not fast enough for mainstream adoption of predictive excellence.

The main obstacles to speed and scale success are the availability and collection of real data on the ground. We find that none of the companies' data is similar to those used in mainstream machine learning, where they record reliable tag information (such as clicks and purchases) with high fidelity and large amounts of records. Many organizations lack the processes and techniques to accurately record key events. Most organizations lack consistent terminology to talk about specific types of operating conditions and failures in complex processes.

Predictive Premium Operations avoids this problem by collecting information about the failure context in real time. Notify the factory of early warning of adverse events and analyze the root causes of these warnings according to the patterns found in the data. Typically, you can do this even before the event affects its operation. Because these insights are derived from current operational data rather than from the analysis of old data, plant engineers are more likely to validate and benefit from their findings. This creates the first time value needed for digital manufacturing organizations.

In this new way, we are rethinking operational AI at Falkonry: no data scientist, data engineer, system integrator. This means that the factory can connect to the AI system within hours. Insights can be provided before problems begin to affect production. Within a few days, plant engineers began to derive value from operational AI. Through the learning system to get feedback from factory engineers, the learning system can immediately improve the opinion generation process. This approach can be applied to dozens of different use cases at the same time, and value can be created more quickly in each use case. The marginal cost of predictive analysis in this method is basically zero.

Over the past three years, we have worked across 12 industries, including manufacturing and national security, with more than 60 customers in more than 60 use cases. The main lesson we have learned from this is that learning systems can generate insight from very little historical data to almost no historical data, and get expert feedback effortlessly. This is the only way customers can deploy and benefit from a wide range of use cases in the enterprise. For a short time.

The value of reducing the time to first deployment will remove many other obstacles required for scale deployment. Plant managers can implement this approach without a large amount of company resources, plant engineers can find insights and improve production, and manufacturing supervisors can manage the process of recording and utilizing expert knowledge.

This is the most advanced technology, and the way we are at Falkonry is changing the dialogue around digital manufacturing. In the process of our in-depth transition, please pay attention to this space, and will share our findings with you, more information will be found in the vibrating chain.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.